CoolFace
Modelpublic

xenobordom/dialogpt-isa-trained-1755493402

sourceHugging Faceapache-2.0updated 1y agoView on Hugging Face
0likes
Model Card

dialogpt-isa-trained-1755493402

通过ISA Model SDK训练的中文对话模型，基于 microsoft/DialoGPT-small 使用 LoRA 方法微调。

模型详情

  • —基础模型: microsoft/DialoGPT-small
  • —训练方法: LoRA Fine-tuning
  • —训练数据: 自定义中文对话数据集
  • —框架: ISA Model SDK
  • —训练损失: 36.2578
  • —可训练参数: 294,912 (0.24%)

使用方法

python
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("xenobordom/dialogpt-isa-trained-1755493402")
model = AutoModelForCausalLM.from_pretrained("xenobordom/dialogpt-isa-trained-1755493402")

# 生成对话
inputs = tokenizer("你好", return_tensors="pt")
outputs = model.generate(**inputs, max_length=50, do_sample=True, temperature=0.7)
response = tokenizer.decode(outputs[0], skip_special_tokens=True)
print(response)

训练配置

  • —Epochs: 2
  • —Batch Size: 1
  • —Learning Rate: 5e-05
  • —LoRA Rank: 8
  • —LoRA Alpha: 16

测试生成

输入: "你好" 输出: "你好!"

训练于 2025-08-18T05:03:59.353827 使用 ISA Model SDK